AWS CodeCommit - Git Hosting That AWS Quietly Killed

AWS killed CodeCommit on July 25, 2024 with about as much ceremony as shutting down a broken vending machine. They just quietly announced it in a blog post that basically said "thanks for playing, now fuck off to GitHub."

What This Means If You're Stuck on CodeCommit

Your existing repos still work, but AWS isn't fixing anything new that breaks. They'll keep the lights on and patch security vulnerabilities, but don't expect any improvements. You're basically using a zombie service that AWS is slowly starving to death.

The real kick in the teeth? CodeCommit was actually the one AWS developer tool that didn't suck. The IAM integration worked properly, you could control access at the repo level with resource-based policies, and it played nice with other AWS services. Now you get to migrate to GitHub's team-based permission model where you'll spend weeks figuring out how to replicate the same granular access patterns.

Why AWS Really Killed It

AWS is getting out of the developer tools business because they kept getting their ass kicked by GitHub's market dominance and GitLab's enterprise features. Instead of competing, they decided to focus on infrastructure services where they actually make money. They killed eight developer tools products in one go, which tells you everything about how much they cared about developer experience.

We found out about the deprecation when one of our developers tried to set up a new repo and got a cryptic 403 Forbidden error. No email warning, no AWS Health Dashboard notification. Just a 403 error that took 30 minutes to figure out meant "no new customers allowed." The error codes documentation doesn't even mention this scenario.

Your Migration Timeline (Spoiler: Start Now)

AWS hasn't announced when they're pulling the plug completely, but waiting around is stupid. Start migrating now while you can take your time instead of scrambling when they announce a sunset date. Their migration documentation is decent, but you'll still need to budget time for:

• Migrating all your repos (obvious, but 50+ repos takes longer than you think)
• Rebuilding your CI/CD pipelines to work with the new Git host (we had 23 CodePipeline configs break)
• Retraining your team on the new platform (budget 2 weeks minimum)
• Fixing all the IAM permissions that won't work anymore (this took us 3 days of head-scratching)
• Updating your CodeBuild projects to use new sources (every single one needs touching)
• Reconfiguring CodeDeploy applications (webhook setup is a pain in the ass)
• Testing all your CloudFormation templates that reference CodeCommit (spoiler: they'll break)
• Updating documentation and runbooks with new Git URLs (we missed 12 places the first time)
• Setting up new SSH keys or access tokens (because GitHub's permission system is totally different)
• Training team on GitHub Actions or GitLab CI/CD (steeper learning curve than expected)
• Reviewing and updating AWS CloudTrail logs for CodeCommit access
• Planning new backup strategies since you're leaving AWS ecosystem (GitHub's backup story sucks)

The One Thing That Didn't Suck: IAM Integration

CodeCommit's IAM integration was legitimately good. You could control who accessed what repos, down to individual branches and files using conditional statements. No more managing SSH keys or dealing with GitHub's confusing permission system. You just used the same IAM roles and policies you already had set up.

The encryption was solid too - HTTPS and SSH in transit, KMS encryption at rest if you wanted it. For compliance-heavy environments, this was actually a selling point - SOC, ISO 27001, and FedRAMP certifications out of the box.

AWS Integration That Actually Worked

If you were all-in on AWS, CodeCommit played nice with CodePipeline, CodeBuild, and CodeDeploy. No webhooks to configure, no external integrations to break. Push to CodeCommit, pipeline starts automatically. It was simple and it worked.

The Pricing Was Actually Fair

$1 per user per month after the first 5 free users. No bullshit about public vs private repos, no weird storage limits that kick in at random sizes. You paid for users, you got unlimited repos. For small teams, it was effectively free.

What CodeCommit Couldn't Do (The Big Problem)

No pull requests. No issue tracking. No code review tools. No project management. No wikis. No nothing beyond basic Git hosting.

You had to bolt on GitHub for pull requests, Jira for issues, and some other tool for code reviews. It was like AWS gave you a really nice garage but forgot to include the car.

This is why nobody outside of hardcore AWS shops used it. GitHub and GitLab give you everything in one place. CodeCommit gave you really good Git hosting and told you to figure out the rest yourself.

The one thing that will bite you in the ass: CodeCommit's IAM policies used resource-based permissions that don't translate to GitHub's team-based model. You'll find yourself manually recreating access patterns that were three lines of IAM JSON.

What You'll Miss After Migration

When you migrate off CodeCommit, you're going to miss a few things (and spend weeks trying to replicate them in your new platform):

• The seamless IAM integration that just worked
• Cross-account access using AssumeRole
• VPC endpoints for private network access
• Integration with AWS Config for compliance tracking
• CloudWatch metrics and monitoring built-in
• AWS X-Ray tracing for Git operations
• Native AWS CLI support without extra setup
• Resource-based policies for fine-grained access
• AWS Security Token Service integration
• AWS Organizations service control policies
• AWS SSO integration that actually worked